nV News Forums

 
 

nV News Forums (http://www.nvnews.net/vbulletin/index.php)
-   NVIDIA Linux (http://www.nvnews.net/vbulletin/forumdisplay.php?f=14)
-   -   Myth recording video breaks up with VDPAU (http://www.nvnews.net/vbulletin/showthread.php?t=139911)

MarkG 10-11-09 04:50 PM

Myth recording video breaks up with VDPAU
 
Some Mythtv ATSC HD recordings from the CBS have video breakups when played back using VDPAU. I'm using mythtv 0.21_0-0.21-212.fc10_bijou20090831 from ATrpms that I built myself against 190.36 on F10 x86_64. They play correctly when using software decoding.

I've uploaded a clipped sample 0testbad.mpg to the partners site. The video breaks up at about 6 and 12 seconds into the clip. I'm using inverse telecine and advanced 2x deinterlacing. This has happened with several different configurations and versions of Myth and Nvidia drivers. Let me know if more info is required.

Mark

MarkG 10-11-09 10:58 PM

Re: Myth recording video breaks up with VDPAU
 
Quote:

Originally Posted by MarkG (Post 2102651)
I've uploaded a clipped sample 0testbad.mpg to the partners site. The video breaks up at about 6 and 12 seconds into the clip.
Mark

When playing with Mplayer SVN-r29770-4.1.2 the video breaks up at 6 seconds using VDPAU and looks OK using software decoding. When software decoding is used, I get these to the console:

At 6 sec:
concealing 120 DC, 120 AC, 120 MV errors

At 12 sec:
concealing 60 DC, 60 AC, 60 MV errors

Mark

Stephen Warren 10-13-09 10:52 AM

Re: Myth recording video breaks up with VDPAU
 
The errors from MPlayer imply that the stream was corrupted during transmission/recording. MPlayer's error concealment may be better than VDPAU's for this clip on some HW.

Can you please supply an nvidia-bug-report. Thanks.

MarkG 10-13-09 07:47 PM

Re: Myth recording video breaks up with VDPAU
 
1 Attachment(s)
I guessed that the file was corrupt in some way and that ffmpeg could handle it differently than vdpau. I would like to determine if the corruption is due to bad reception or the broadcaster's encoder having an issue. If it is the latter, there is nothing I can do to improve things, and I would hope that you could get VDPAU to handle it better. Using ffmpeg there is no obvious visual impairment.

Mark

Stephen Warren 10-16-09 11:22 AM

Re: Myth recording video breaks up with VDPAU
 
Thanks for the information. We'll look into it.

I doubt we'll be able to tell why the clip is corrupted.

MarkG 10-16-09 12:37 PM

Re: Myth recording video breaks up with VDPAU
 
Quote:

Originally Posted by Stephen Warren (Post 2105546)
Thanks for the information. We'll look into it.
I doubt we'll be able to tell why the clip is corrupted.

Thanks.

If not, hopefully you may be able to improve playback of this type of corruption. Ffmpeg is able to "conceal" the errors so visually it looks fine. With VDPAU, shows like this either break up into blocks or stutter several times a minute.

Mark

MarkG 11-29-09 04:49 PM

Re: Myth recording video breaks up with VDPAU
 
195.22 has the same corruption on playback.

Have you been able to investigate the issue?

Mark

Stephen Warren 11-30-09 11:43 AM

Re: Myth recording video breaks up with VDPAU
 
Sorry, not yet.

MarkG 01-25-10 12:56 PM

Re: Myth recording video breaks up with VDPAU
 
195.30 has the same issue. Is there any chance a newer card may be able to decode this better, say a GT240? It seems that ffmpeg can guess on the missing motion vectors and "conceal" the problem. If the newer cards have a better h.264 decoding algorithm in the hardware, that may be an option.

Mark

cehoyos 01-26-10 05:21 AM

Re: Myth recording video breaks up with VDPAU
 
Quote:

Originally Posted by MarkG (Post 2172536)
195.30 has the same issue. Is there any chance a newer card may be able to decode this better, say a GT240? It seems that ffmpeg can guess on the missing motion vectors and "conceal" the problem. If the newer cards have a better h.264 decoding algorithm in the hardware, that may be an option.

Note that FFmpeg has been known for a long time for its advanced error-concealment. I may be wrong, but I think it is unlikely to implement something similar in a hardware decoder.
You should really try to improve your reception, did you also try to record with szap and cat, just in case it is not a reception probem?

Carl Eugen

MarkG 01-26-10 12:46 PM

Re: Myth recording video breaks up with VDPAU
 
1 Attachment(s)
Quote:

Originally Posted by cehoyos (Post 2172945)
Note that FFmpeg has been known for a long time for its advanced error-concealment. I may be wrong, but I think it is unlikely to implement something similar in a hardware decoder.
You should really try to improve your reception, did you also try to record with szap and cat, just in case it is not a reception probem?

Carl Eugen

I've got a rooftop antenna with line of sight to the transmitters, with the best HD frequency only, sub 2 dB noise figure, high overload amp i could find. My tuner cards don't report accurately, but they show zero bit error rate and the TV that is also split off the feed shows the high 90s for signal strength. I think the reception is as good as I can get it.

I am concerned that they now allow adjacent channels for digital stations. The station with the problem does happen to have an occupied adjacent channel. I've attached the output of a spectrum analyzer. The problem station is the highest frequency one on the right. It looks strong and well above the noise to me, but my analyzer can't evaluate the bit error rate of ATSC signals. I can't tell if the problem is a reception problem or too short a set of errors to not be corrected by the ECC. Most reception problems I've seen are larger and cause breakup in both the picture and sound.

I uploaded a sample of the bad video to Nvidia in the hopes that someone with an mpeg bitstream analyzer could tell. I am thinking that it may be possible that the station's encoder is not quite right and it is just covered up by most receivers and the rest just accept it. I get the same issues with any of my three tuner cards, which are two different models. I've not tried recording with szap and cat.

Mark

cehoyos 01-27-10 04:57 AM

Re: Myth recording video breaks up with VDPAU
 
Quote:

Originally Posted by MarkG (Post 2173251)
I've got a rooftop antenna with line of sight to the transmitters, with the best HD frequency only, sub 2 dB noise figure, high overload amp i could find. My tuner cards don't report accurately, but they show zero bit error rate and the TV that is also split off the feed shows the high 90s for signal strength. I think the reception is as good as I can get it.

That is the reason why I would recommend to try recording with szap and cat, possibly looking at femon (and see if the problem is reproducible).

Carl Eugen


All times are GMT -5. The time now is 11:37 AM.

Powered by vBulletin® Version 3.7.1
Copyright ©2000 - 2014, Jelsoft Enterprises Ltd.
Copyright 1998 - 2014, nV News.